Optimove Named a “Strong Performer” in the Cross-Channel Campaign Management Forrester Wave | Read More

Search the website

Behavior-based Customer Segmentation for More Effective Retail Marketing

Retail Customer Segmentation

Retail marketers are constantly looking for ways to improve the effectiveness of their campaigns. One way to do this is to target customers with the particular offers most likely to attract them back to the store and to spend more on their next visit. In other words, the marketer’s goal is to make the most relevant match between customer and offer. This article presents a highly effective means of accomplishing this goal using shopping behavior-based customer segmentation.

The Marketing Benefits of Customer Segmentation

To get the most uplift from a marketing campaign, the campaign should be directed at the customers most likely to respond to it. By using customer segmentation to determine actionable “customer prototypes,” the marketer is able to test different campaigns on particularly relevant target groups of customers. Over time, the results of these campaigns can be measured and compared to find the most effective repeatable offers to each customer segment.

For example, once a segment of supermarket customers is identified as being “meat lovers,” the marketer could try a variety of up-sell campaigns (e.g., to encourage customers to buy different, more expensive kinds of meats) and cross-sell campaigns (e.g., to encourage customers to buy products from other categories, such as gourmet meat sauces or barbecue accessories). Besides short-term sales, this approach typically increases long-term customer loyalty as well.

Customer Segmentation Using Cluster Analysis

In brief, cluster analysis uses a mathematical model to discover groups of similar customers based on finding the smallest variations among customers within each group. The process is not based on any predetermined thresholds or rules (as are most simple segmentation methods), but rather the data itself generates the customer prototypes that inherently exist within the population of customers.

The two main advantages of cluster analysis over simple threshold/rule-based segmentation are (1) practicality – it would be practically impossible to use predetermined rules to segment customers over many dimensions, and (2) homogeneity – variances within each resulting group are very small in cluster analysis, whereas rule-based segmentation typically groups customers who are actually very different from one another.

How to Perform Cluster Analysis

The first step in cluster analysis is to prepare the customer spend data for each product category. Grocery stores and supermarkets would typically look at categories such as packaged foods, meat, dairy, produce, seafood and bakery. More granular category levels can also be selected if the goal is to segment customers within a particular known group. For example, the supermarket could try to market specialty cheeses to cheese lovers by segmenting customers based on their purchases of various categories of cheese (e.g., Cheddar, Cottage, Monterrey Jack, Mozzarella, Swiss).

The next step is to perform cluster analysis on all customers to identify distinct homogeneous groups of customers with minimal variance between their purchasing behaviors. This identifies unique “customer prototypes” (such as meat lovers, produce lovers and gourmet lovers) to which specific marketing offers can be targeted.

(It is important to mention that part of the cluster analysis process is normalizing price levels across departments using weights related to the average basket price in each department. This prevents the inherent differences in the absolute prices between departments from skewing the analysis.)

The following chart shows partial results of a cluster analysis performed by a chain of grocery stores:

This chart shows a seven-dimension cluster analysis which resulted in the discovery of five customer prototypes: basic shoppers (who purchase a variety of products, but mostly packaged items), meat lovers (who purchase a large amount of meat), produce lovers (who purchase a large amount of fresh produce), gourmet lovers (who purchase mostly items from the gourmet department) and variety shoppers (whose shopping behavior is widely spread among departments).

Marketing to the Customer Prototypes

Once the marketer has a clear view of the various customer prototypes, it makes sense to target relevant marketing campaigns to the most interesting segments. An integral part of testing the effectiveness of these campaigns is to divide the prototype group into a “test group” and a “control group” and then to compare the uplift between them.

By constantly experimenting, measuring and improving, it becomes possible to discover the most effective combinations of marketing campaign and customer segment.

The Required Tools

The heart of the approach described in this article is cluster analysis. Cluster analysis, usually based on the k-means algorithm, is not something new. Many statistical software packages provide the means to perform cluster analysis and there are even Excel add-ins available for the purpose.

Retailers without an in-house team of statisticians, however, may want to look at marketing-oriented software that incorporate cluster analysis as part of an overall customer analysis and marketing campaign management application. The better applications in this category will also include marketing-oriented features such as dynamic customer group generation, automatic selection of test and control groups and automatic measurement and reporting of campaign results.

Segmentor
Comments (2)
Enter your comment
Enter a valid email address
Submitting comment...
  • Avatar
    Tal

    Well, I know that usually no one ever comments on blog posts (so I don’t know if there would be an answer), but I really enjoyed this post and wanted to note that.
    Also, I wanted to make a comment about price levels, for which you suggested normalization etc..
    I would suggest that instead of normalizing relative pricing data – using absolute quantities for each department/category.
    Two of the reasons that I could think of right now –
    – As you said, the inherited difference between products is something to consider
    – Being exposed to inflation/deflation/price flactuation
    Obviously, it is case/client/question specific, but generally, this should provide, I believe, a faster, cleaner analysis.

    What do you say?

    Thank you,

    • Avatar
      Jan

      Hi Tal,

      Sounds like an interesting approach, but this also generates a bias which favours high-frequency categories. Some categories are bought weekly (e.g. vegetables) while others are only bought once every few months (e.g. deodorants).

      So volume differences in vegetables tend to be much bigger than in deodorants.

      Therefore I would suggest to use category volume or spend relative to the category heavy buyers. For instance, if the 95th percentile spends 500$ per year on a specific category and you spend 200$ you would be at 40%. This percentage will represent per category how you “stack up” to the heavy buyers and how intensively you are buying then. So it can clearly show preferences.

Comments (2)
Leave a reply
Cancel
Enter your comment
Enter a valid email address
Submitting comment...

Stay in touch

Be the first to know all about the latest Marketing tips & tricks, Industry special insights and more

WP Feedback

Dive straight into the feedback!
Login below and you can start commenting using your own user instantly